Recent studyBlind TTS Elo is live. Compare two anonymous voice samples, vote after listening, and help separate real preference signal from noise.Vote in the study ->
Text-to-speech · Polish hard-text

Polish TTS needs entity exactness, not just pretty speech.

The Polish wedge stresses diacritics, inflected text, dates, money, addresses, order IDs, and abbreviations. A model that sounds fluent but drops zł, św., or product-code characters should fail this track.

Measured leaderboardAPI example
Seed prompts
  1. Zażółć gęślą jaźń.
  2. Pchnąć w tę łódź jeża lub ośm skrzyń fig.
  3. Spotkanie odbedzie sie 17.05.2026 o 14:45.
  4. Koszt wynosi 1 234,56 zł brutto.
  5. Wyślij maila na kacper.żółć@example.pl.
  6. Nr zamówienia: PL-2026/05/17-AŹ-09.
  7. ul. św. Wojciecha 12/4, 61-749 Poznań.
Track

hardtext-pl

Polish diacritics, dates, currencies, addresses, abbreviations

Metrics
Polish CER with diacritics
number/date/currency exactness
named-entity accuracy
abbreviation handling: ul., św., dr, mgr, zł
ASR transcript diff